DR setup in VMWare

I have a network as below all hosts in Vcenter cluster. DR and production is connected by fiber as of now.
EE.pngIf I change as below is it possible to failover to DR automatic. Changes in orange color.
 EE---Copy.png
Your advise is appreciated.
LVL 29
MAS (MVE)Technical Department HeadAsked:
Who is Participating?
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
You've got options then if you now have SRM use it.

SRM is usually used to a Production and DR Site and failover and fallback to Prod

If you have A Stretched VLAN between Prod and DR no SRM easier implementation
0
 
andyalderCommented:
It's unclear whether your SAN switches are fibre channel or iSCSI but either way adding a second switch won't affect failover since a single switch can be logically split in two by zoning or vLANs. Adding physical links at the SAN level can allow direct storage to storage replication rather than doing it across Ethernet but that just ensures two copies of the data which I'm sure you have already.

Automating failover isn't trivial and really has to be tackled at the service level rather than hardware; a simple hardware solution would move the VM that controlled the building access gate to the DR site and as the inter-site link is down the repair men couldn't get in to fix the fault. The 5000 user Exchange server starts at DR site but only a few users need it as their desks are on the main site which is down so they took the day off. Failback is even harder to automate.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
Is it vmware or hyperv?

if vmware first but even for hyperv(if so change to vmware) we need specific configurations in vcenter or hyperv manager and network connectivity as the buildings and infrastructure far apart. IE VPN, or direct hgh speed link,

We need max 1MS link between the two sites to failover lets say a 75TB replicated 10GB SAS SAN locally , 1GB network wise to do an instant failover, these are fairly solid rough stats from my experience.

What we need is VMWARE SRM on periodic sync of lets say 15/30 minutes if your business can afford it, non instant failover manually initiated.

#1
0
Network Scalability - Handle Complex Environments

Monitor your entire network from a single platform. Free 30 Day Trial Now!

 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
It states VMware in OP
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
If you have replicated SAN LUNs you can automate failover with SRM

No SRM then it can be done manually
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
Its VMWare and all hosts part of VCenter cluster.
As of now all are 1G fiber. We are planning to upgrade all NICs to 10G if the second plan works.

So we need storage replication and SRM?
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
Correct for Automatic!

Manual you could do yourself

It depends really how automatic you want it versus cost of SRM

Or you could look at Veeam Nakivo or Zerto
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
If DR hosts could see Production SAN

VMware HA would work
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
If both switches connected each other as in the second image above DR Hosts will see SAN.
correct me if I am wrong
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
So if all networking is in place e.g. DR site has same networking VLANs as Production or what is sometimes called stretched VLAN between Prod and DR

DR hosts have access to Prod datastore s and are mounted

In the event of Prod failing VMs will restart at DR using VMware HA

No SRM required
1
 
MAS (MVE)Technical Department HeadAuthor Commented:
Is  there any special settings required to not move VMs automatically to DR hosts when production hosts are active?
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
the vms are held on the SANS and kept up to dste here. the biggest challenge is getting them in sync locally on site.

On site activation one of the LUNs is locked and one is unlocked.

My specialty is periodic which uses SRM, I have looked at specs for stretch cluster and was told to be careful RE 1ms latency which is why i posted you that network spec.

What SANS you runnin?


andrew you will forgive me for not seeing SRM i hope :P

Also 'RE 1GB(down to 15 or even 5min periodic SRM) vs 10GB(stretch cluster)

networking is critical here
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
If you have DRS create a Production and DR group and rules to keep VMsbin the correct groups
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
Reason we need 1MS ping here before going instant sync and not periodic is because every MS counts times many many many times in this kind of setup.

Genuinely if your using vm servers with some decent IOs 1GB direct connection pipe via same ISP is probably not even enough.

Stay away from traffic shaping and dedicate your whole link for this.

1
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
In case of failure of host in Prod how much time it will take to move/start a VM to DR host?
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
1-2mins for HA maybe less
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
VMs are restarted so loss of service for 1-2mins
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
i guess you gotta decide whether you want periodic or instant.

Decision comes down to what the business recovery time is.

in my case fifteen mins with good ttl 15 min again was perfect.

i thought be careful of what you wish for in regards to instant. we wanted more control over our failovers.
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
Is it better to have separate VLANs for Prod and configure SRM?
Or better to have stretched VLAN and configure HA?
They have SRM license now.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
depends if your connecting the sites by vpn your going to have seperate vlans well it will make no difference of they are some way linked at switch level across the vpn or not

you will have 2 x fw running the connections between both locations

where as if we have direct link via same isp. essentially a high speed dedicated switch port between two sites you can network that connection off on both sides behind a fw if you like and do what you want with it in either site.

I actually prefer just opening out that direct isp port for the replication in both sites by spreading the connection into a switch which only the replicated nics are connrcted too, simple works well but check pci on that if applicable.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
as your using direct fiber it hardly matters imo. just have all hosts in same vlan ID
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
as mentioned previously i would be very wary of a instant setup with 1gb its borderline at best unless your running fairly static servers
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
I think you now have all your answers!

Other than HOW to implement it!

Which is a very different question!
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
SRM actually operates of a protected site A and recovery site B this is done by LUN replication direction at back end.

We can also do Bi Directional and run some servers in A and some in B if we like and SAN supports it!
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
What do you think about some potential networking setups Andrew? :P
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
You also need check SRA compatibility as what SRM is doing is a HTTPS API call lock to SAN via VMWARE
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
As of now we have HP 3PAR on both site. Can configure storage replication as well.
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
SRM will use it!
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
@Andrew Hancock
With stretched VLAN or separate VLAN for SAN network?

if we use SRM VMs will restart in DR host or live migrate?
What is the chance of data loss in my case with storage replication ?
We need another Vcenter server in DR if we configure storage replication and configure SRM?
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
What are you trying to implement because your questions are now confusing HA and SRM?
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
We are trying to make zero downtime or least downtime possible when a host fails in production with the current resources we have. And DR is in the same geographical location.  i.e. next building.
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
In the event of Production failing if a bomb was to hit how an earth can VMs be Live Migrated!

Think about it all VMs are Down and Gone!

Therefore SRM Starts Up VMs from replicated storage

The VMs are as old as last replicate cycle!

Yes another vCenter is required and you will have to redesign current environment to have a Production and real DR Site
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
Zero downtime use FT
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
No chance of data loss
0
 
Andrew Hancock (VMware vExpert / EE MVE^2)VMware and Virtualization ConsultantCommented:
Again your design for SAN depends if you do SRM or HA and what networking can be achieved

Some organisations cannot do stretches VLAN

If you can use it makes things much easiet
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
at 3par level RCIP ports run on HTTPS protocol. they have CRC built into the traffic exchanges etc to ensure you cannot lose data.

Create a new VV in management console. make it replicated. use it for UAT and test VMs. have a look at 3PARUG forum and user markbills post history. we learned and outlined a lot of stuff here. best of luck.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
and regarding design 3par has few ways to do this.

you can actially do rcip streaming for instant failover, rcip is usually only for periodic plus SRM

usually you will need 3rd site for witness quorum if you are switching to full sync top tier solution.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
3par is very specialized technology make no mistake.
0
 
Mark BillExchange, AD, SQL, VMware, HPE, 3PAR, FUD, Anti MS Tekhnet, Pro EE, #1Commented:
no offence but by the sounds if it you need 3par consultancy or the manual. Of course i can help in the right circumstances.
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
Sorry to object.
0
 
MAS (MVE)Technical Department HeadAuthor Commented:
Thanks to Andrew and Mark.
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.